基于BP神经网络的星箭界面动载荷识别  

Dynamic load identification of satellite-rocket interface based on BP neural network

作　　者：陈树海 郭安丰 吴邵庆[2] 费庆国[1] CHEN Shuhai;GUO Anfeng;WU Shaoqing;FEI Qingguo(School of Mechanical Engineering,Southeast University,Nanjing 211189,China;School of Civil Engineering,Southeast University,Nanjing 211189,China;Shanghai Satellite Engineering Research Institute,Shanghai 201109,China)

机构地区：[1]东南大学机械工程学院,南京211189 [2]东南大学土木工程学院,南京211189 [3]上海卫星工程研究所,上海201109

出　　处：《振动与冲击》2023年第5期279-286,304,共9页Journal of Vibration and Shock

基　　金：江苏省优秀青年基金(BK20180062);江苏省“六大人才高峰”项目(KTHY-005)。

摘　　要：提出一种基于BP神经网络的星箭界面动载荷识别新方法。建立卫星结构的高保真动力学模型,利用仿真分析/地面试验获取的卫星结构加速度响应与星箭界面加速度激励的样本库;基于BP神经网络训练卫星结构加速度与星箭界面加速度激励的传递关系,利用实测卫星结构加速度响应识别星箭界面加速度激励;将星箭界面加速度激励施加与卫星结构的高保真动力学模型获取星箭界面动载荷。开展了数值仿真和振动试验,验证了所提出方法的有效性,为服役状态下卫星结构的振动载荷环境预示提供有力支撑。Here, a new method based on BP neural network to identify dynamic load of satellite-rocket interface was proposed. Firstly, the high-fidelity dynamic model of satellite structure was established, and the sample library of satellite structure acceleration response and satellite-rocket interface acceleration excitation obtained using simulation analysis/ground tests was adopted. Secondly, BP neural network was used to train the transfer relation between satellite structure acceleration and satellite-rocket interface acceleration excitation, and the measured satellite structure acceleration responses were used to identify satellite-rocket interface acceleration excitation. Finally, the acceleration excitation of satellite-rocket interface was exerted on the high-fidelity dynamic model of satellite structure to achieve dynamic load of satellite-rocket interface. Numerical simulation and vibration tests were conducted to verify the effectiveness of the proposed method. It was shown that the proposed method can provide a strong support for predicting vibration load environment of satellite structure in service.

关 键 词：星箭界面 加速度激励 界面力识别 BP神经网络 试验验证 

分 类 号：TB123[理学—工程力学]